Manuel Feuchter
Investigations on Joule heating applications by multiphysical continuum simulations in nanoscale systems
Schriftenreihe des Instituts für Angewandte Materialien Band 43 Investigations on Joule heating applications in nanoscale systems M. Feuchter
43
Manuel Feuchter
Investigations on Joule heating applications by multiphysical continuum simulations in nanoscale systems Schriftenreihe des Instituts für Angewandte Materialien Band 43
Karlsruher Institut für Technologie (KIT) Institut für Angewandte Materialien (IAM)
Eine Übersicht aller bisher in dieser Schriftenreihe erschienenen Bände finden Sie am Ende des Buches. Investigations on Joule heating applications by multiphysical continuum simulations in nanoscale systems by Manuel Feuchter Dissertation, Karlsruher Institut für Technologie (KIT) Fakultät für Maschinenbau Tag der mündlichen Prüfung: 08.07.2014
Impressum
Karlsruher Institut für Technologie (KIT) KIT Scientific Publishing Straße am Forum 2 D-76131 Karlsruhe KIT Scientific Publishing is a registered trademark of Karlsruhe Institute of Technology. Reprint using the book cover is not allowed. www.ksp.kit.edu
This document – excluding the cover – is licensed under the Creative Commons Attribution-Share Alike 3.0 DE License (CC BY-SA 3.0 DE): http://creativecommons.org/licenses/by-sa/3.0/de/ The cover page is licensed under the Creative Commons Attribution-No Derivatives 3.0 DE License (CC BY-ND 3.0 DE): http://creativecommons.org/licenses/by-nd/3.0/de/
Print on Demand 2014 ISSN 2192-9963 ISBN 978-3-7315-0261-6 DOI 10.5445/KSP/1000042982
Investigations on Joule heating applications by multiphysical continuum simulations in nanoscale systems
Zur Erlangung des akademischen Grades
Doktor der Ingenieurwissenschaften
der Fakultat¨ fur¨ Maschinenbau
Karlsruher Institut fur¨ Technologie (KIT)
genehmigte
DISSERTATION
von
Dipl.–Ing. Manuel Klaus Ludwig Feuchter geboren am 20.11.1983
in Wertheim
Tag der mundlichen¨ Prufung:¨ 08.07.2014 Hauptreferent: Prof. Dr.–Ing. M. Kamlah Korreferent: Prof. Dr. rer. nat. C. Jooss Korreferent: Prof. Dr. rer. nat. O. Kraft
‘No subject has more extensive relations with the progress of industry and the natural sciences; for the action of heat is always present, it influences the processes of the arts, and occurs in all the phenomena of the universe.’ – Jean Baptiste Joseph Fourier [cf. 118, p. 241]
Abstract
A requirement for future thermoelectric applications are poor heat conducting materials. Nowadays, several nanoscale approaches are used to decrease the thermal conductivity of this material class. A promising approach applies thin multilayer structures, composed of known materials. Here, the nanoscale stacking affects the heat conduction and provokes new emergent thermal properties; However, the heat conduction through these materials is not well understood yet. Therefore, a reliable measurement technique is required to measure and understand the heat propagation through these materials. In this work, the so-called 3ω-method is focused upon to investigate thin strontium titanate (STO) and praseodymium calcium manganite (PCMO) layer materials. Previously unexamined macroscopic influence factors within a 3ω-measurement are considered in this thesis by Finite Element simulations. Thus, this work furthers the overall understanding of a 3ω-measurement, and allows precise thermal conductivity determinations. Moreover, new measuring configurations are developed to determine isotropic and anisotropic thermal conductivities of samples from the micro- to nanoscale. Since no analytic solutions are available for these configurations, a new evaluation methodology is presented to determine emergent thermal conductivities by Finite Element simulations and Neural Networks.
v
Kurzfassung
Zukunftige¨ thermoelektrische Anwendungen erfordern thermisch schlecht leitende Materialien. Hierfur¨ werden heutzutage verschiedene Ansatze¨ verfolgt um die Warmeleitf¨ ahigkeitszahl¨ dieser Materialklasse zu verringern. Ein vielversprechender Ansatz verwendet dunne¨ Vielschichtstrukturen die sich aus bekannten Materialien zusammen setzen. Hier wird durch nanoskaliges Schichten die Warmeleitung¨ beeinflußt und es werden neue thermische Eigenschaften hervorgerufen. Jedoch ist die Warmeleitung¨ durch solch ein Material bis heute noch nicht bis ins Detail verstanden. Zu diesem Zweck bedarf es einer verlaßlichen¨ Messmethode, um die Warmeausbreitung¨ durch solch ein Material messen und somit verstehen zu konnen.¨ Diese Arbeit konzentriert sich auf die so genannte 3ω-Methode zur Erforschung dunner¨ Strontium Titanat (STO) und Praseodym Calcium Manganit (PCMO) Schichtmaterialien. Bisher unbeachtete makroskopische Einflußfaktoren, die wahrend¨ einer 3ω-Messung auftreten, werden in dieser Arbeit durch Finite Element Simulationen berucksichtigt.¨ Dadurch tragt¨ dieses Werk zum Gesamtverstandnis¨ einer 3ω-Messung bei und erlaubt eine prazise¨ Bestimmung der Warmeleitf¨ ahigkeitszahl.¨ Daruber¨ hinaus werden neue Messkonfigurationen zur Bestimmung der isotropen als auch anisotropen Warmeleitf¨ ahigkeitszahl¨ fur¨ mikro- bis nanoskalige Proben entwickelt. Da fur¨ diese Messkonfigurationen keine analytischen Losungen¨ verfugbar¨ sind, wird eine neue Methodik zur Auswertung und Bestimmung der Warmeleitf¨ ahigkeitszahl¨ vorgestellt, die Finite Element Simulationen und Neuronale Netze kombiniert.
vii
Contents
Nomenclature xiii
1 Introduction 1 1.1 Thermoelectricity and material selection ...... 8 1.2 Heat conduction in solids ...... 16
2 The 3ω-method 21 2.1 Concept, measurement principle and geometry configurations ...... 21 2.2 Top down geometry ...... 26 2.2.1 Heat source on bulk materials ...... 26 2.2.2 Heat source on layer-substrate materials . . . . 35 2.3 Bottom electrode geometry ...... 40 2.3.1 Heat source between bulk-like materials . . . . 40 2.3.2 Heater substrate platform ...... 42
3 Finite Element Model 45 3.1 Governing equations ...... 46 3.1.1 Temperature field ...... 46 3.1.2 Electromagnetic field ...... 52 3.1.3 Mechanical field ...... 59 3.2 Transient analysis ...... 63 3.3 Eigenfrequency analysis ...... 65 3.4 Mesh-block building system ...... 67
ix 4 Investigation of varying structures and multiphysical couplings 69 4.1 Top down geometry ...... 69 4.1.1 Heat source on bulk materials ...... 69 4.1.1.1 Validation of Finite Element Model . . 69 4.1.1.2 Real substrate size ...... 71 4.1.1.3 Geometrical variations of the heater and its material properties ...... 73 4.1.1.4 General temperature distribution for a heater substrate system ...... 76 4.1.1.5 Temperature dependent resistivity of the heater ...... 78 4.1.1.6 Three-dimensional heat conduction in the substrate ...... 80 4.1.1.7 Radiation at the heater’s surface . . . . 84 4.1.1.8 Surface roughness ...... 86 4.1.1.9 Skin effect in the heater ...... 88 4.1.1.10 Thermal expansion and stresses . . . . 90 4.1.1.11 Eigenfrequency analysis ...... 94 4.1.2 Heat source on layer-substrate materials . . . . 96 4.1.2.1 Monolayer configurations ...... 97 4.1.2.2 Multilayer configurations ...... 104 4.2 Bottom electrode geometry ...... 107 4.2.1 Bulk-like materials ...... 107 4.2.2 Thin layers ...... 111 4.2.3 Application of heat sink ...... 114 4.2.4 Patterned multilayer structures ...... 119 4.3 Membrane structures ...... 121 4.3.1 Two-dimensional models ...... 122
x 4.3.2 Three-dimensional 6-pad heater structure . . . . 127 4.4 Pillar and pad structure ...... 138
5 Methodology to determine the isotropic and anisotropic thermal conductivity 155 5.1 General methodology ...... 155 5.2 Neural Network and Inverse Problem ...... 157
6 Application examples 163 6.1 STO layer on an STO substrate ...... 163 6.2 Bottom electrode geometry ...... 168
7 Summary 175
A Detailed information on formula 181 A.1 Modulated voltage ...... 181 A.2 Information about the approximate solution ...... 182 A.3 Definition of the skin depth and derivation of decoupled equations for the magnetic and electric field intensity .. 185 A.4 Derivation for the complex Helmholtz-Equation . . . . . 188
B Material properties and constants 189
C Technical drawing of the 6-pad heater structure 191
Publications 193
References 197
Acknowledgment - Danksagung 221
xi
Nomenclature
Greek Symbols
α linear temperature coefficient [ 1/K] β linear thermal expansion coefficient [ 1/K]
ΩH area of the heater
Ωlay/m area of the layer or material
Ωs area of the substrate
σs stress tensor Υ amplitude of motion ε infinitesimal strain tensor
εT thermal strain χ, $, ψ, φ complex terms of Borca-Tasciuc’s solution ∆R electric resistance oscillation [ Ω ] B ∆TH temperature amplitude due to Borca-Tasciuc including heater properties [K] B ∆Th temperature amplitude due to Borca-Tasciuc without heater properties [K] C ∆Th temperature amplitude due to Cahill [K] C ∆Tlay temperature amplitude due to layer [K] ∆Tˇ general temperature oscillation [K] ∆T temperature amplitude in the heater/material [ K ] permittivity [ As/Vm ]
0 vacuum constant permittivity [ As/Vm ] relative permittivity of specific material [ ] r − η Neural Network gradient scaling parameter
xiii Γ boundary for the partial differential equation γ Euler-Maschoneri constant [ ] − ΓH1 4 boundary for the magnetic field − ΓInt interface boundary
κeff effective thermal conductivity [W/mK ] κ thermal conductivity [W/mK ]
κe thermal conductivity due to electrons [ W/mK ]
κH thermal conductivity of the heater [W/mK ]
κk, κk+1 thermal conductivities of two contacting materials [W/mK ]
κlay thermal conductivity of the layer [W/mK ]
κmcrs cross-plane thermal conductivity of the investigated material [W/mK ]
κmin in-plane thermal conductivity of the investigated material [W/mK ]
κm thermal conductivity of the investigated material [ W/mK ]
κph thermal conductivity due to phonons [ W/mK ]
κs thermal conductivity of the substrate [W/mK ] ( ) Neural Network regulation term E W ( ) Neural Network error term G W Neural Network activation function Q +1 T Neural Network iterative synaptic weight W T Neural Network actual synaptic weight W Neural Network synaptic weights Wij Neural Network input vector Xj Neural Network output vector Yj µ permeability [ Vs/Am ]
µ0 vacuum constant permeability [ Vs/Am ]
xiv µ relative permeability of specific material [ ] r − ν Poisson’s ratio [ ] − Ω area of the partial differential equation ω angular frequency of the current [ 1/s ] Π Peltier coefficient [W/A] ρ mass density [ kg/m3 ] 2 ρe electric charge density [C/m ] 3 ρm mass density of the investigated material [ kg/m ] σ mechanical stress [N/m2 ]
σb Boltzmann constant [J/K]
σec electrical conductivity [ 1/Ωm ] τ mechanical shear stress [N/m2 ] Θ correction value due to boundary mismatch [ ] − θ angular of rotation [◦ ] ϕ contact potential
%0 specific electric resistivity [ Ωm ]
%0(T) temperature dependent specific electric resistivity [ Ωm ]
ς phase shift [ ◦ ]
℘e electric charge density, complex quantity
Latin Symbols
1/q thermal penetration depth [ m ] f volume force A jinit magnetic vector potential, initial complex quantity
Aj magnetic vector potential, complex quantity [ Vs/m ] B magnetic flux density, complex quantity D electric flux density, complex quantity
xv E electric field intensity, complex quantity H magnetic field intensity, complex quantity
Hmat magnetic field intensity in a certain material J electric current density, complex quantity Fourier cosine transform F B magnetic flux density [ Vs/m2 ] C elasticity tensor D electric flux density [ As/m2 ] E electric field intensity [V/m ] H magnetic field intensity [A/m ] j electric current density [A/m2 ] u mechanical displacement vector A cross-sectional area of the heater [ m2 ] a height of the heater [ m ] 2 Ap circular surface area of the pillar [ m ] b half heater width [ m ] beff effective half heater width [ m ] cpm specific heat capacity of the investigated material [J/kgK ] cp specific heat capacity [J/kgK ] durp pulse duration [ s ] dT/dR specific temperature to resistance behavior [ K/Ω ] dlay thickness of the layer [ m ] dPCMO thickness of pillar PCMO layer [ m ] 2 Ds thermal diffusivity of the substrate [ m /s ] ds thickness of the substrate [ m ] e surface emissivity [ ] − fdp regime length [ s ]
xvi fc frequency of the current [ Hz ] fp pulse repetition rate [ s ] G shear modulus [N/m2 ] htc heat transfer coefficient [W/m2K] i complex number [ ] − I0 peak current [A] B I0 modified Bessel function of first kind and zero order 2 j0 peak current density [A/m ] B K0 modified Bessel function of second kind and zero order L length of the heater [ m ]
Lh(z) length of the heater in z-direction [ m ] n normal to the cross-sectional cut P released power [W] p source term [W/m3 ]
P0 applied power amplitude [W]
P0/L power per length [W/m ] Q amount of heat [W] q heat flux [W/m2 ] R(T) measured electric resistance of the PCMO pillar [ Ω ] R electric resistance [ Ω ] r distance form line source [ m ]
R0 average electric resistance [ Ω ] rp radius of the pillar [ m ] sdp step size [ s ] t time [ s ]
T0 ambient temperature [K]
Tav average temperature rise [K]
Tco constant temperature rise [K]
xvii TH temperature in the heater [K]
Tinit initial temperature [K]
Tk, Tk+1 temperatures of two contacting materials [K]
Tmeanmax maximum mean temperature in the heater [K]
Tmeanmin minimum mean temperature in the heater [K]
Tmean(t) time dependent mean temperature in the heater [ K ]
Tm temperature in the investigated material [K]
Ts temperature in the substrate [K] U voltage
U3ω third harmonic voltage [V] ux,y,z mechanical displacement in the respective spatial direction [ m ] x, y, z spatial coordinates [ m ] 2 Eσ Young’s modulus [N/m ] f arbitrary function k complex quantity of Helmholtz-equation S surface area [ m2 ] I electric current [A] S Seebeck coefficient [ µV/K] T temperature [K] ZT figure of merit for thermoelectrics [ ] −
Miscellaneous
FES Finite Element Simulation SEM scanning electron microscopy
SIN Si3N4
xviii BC boundary condition
PCMO Pr1 xCaxMnO3 − RRAM resistive random access memory
STO SrTiO3
YSZ compound of ZrO2 and Y2O3
PMMA Polymethyl Methacrylate C5H8O2
Operators
divergence ∇ · rotation ∇× gradient ∇
xix
1 Introduction
Motivation By the year 2035, the global energy demand will have increased by one-third of today’s consumption [75]. Although it is controversially discussed from when on primary used energy resources will run short, it is fact that costs for energy sources increased significantly in the recent years and probably will in future [143]. To overcome this dilemma, new technologies for energy production are necessary, accompanied by efficiently using energy resources. In many cases, heat engines are used to convert primary energy sources1 into mechanic and electric energy. However, a tremendous amount of energy is lost by waste heat [cf. 158]. Here, thermoelectric generators can contribute to increase the overall efficiency by turning waste heat into electricity. A typical application example are conventionally driven cars Fig.1(a). Hot exhaust gases pass off into the environment without any use. Here, the car’s overall efficiency could benefit from waste heat recovery. However, besides the enhancement of existing heat engines, thermoelectric generators can be used also at smaller scales in mini devices for energy harvesting. Conceivable applications are embedded electronic devices into human clothing such as sensors, mobile phones or media players Fig.1(b). Application areas that don’t effect daily life of general public are special stand-alone energy systems. Facing the challenge of constant energy support over several years, some satellites Fig.1(c) use radioisotope thermoelectric generators since the early 1960s [129]. This kind of generator is also used in the famous self-sustaining space rover Curiosity on Mars Fig.1(d).
1e.g. oil, coal and gas
1 1 Introduction
Based on the working principle, thermoelectric converters can also be used to cool devices. In recent years, a trend towards electro mobility emerged. The desired operating distance and the apparent mass of the vehicle requires high performance batteries. These batteries emit a significant amount of heat in comparison to small electronic devices when discharged. Here, a cooling regulation might be needed while the car is operated Fig.1(e). Less spectacular, but prevalent are portable refrigerators. (a) (b)
(c) (d)
(e)
Figure 1: Examples for thermoelectric generators: (a) application at an exhaust [65]; (b) mini devices [153]; (c) stand-alone satellites [135]; (d) self-sustaining space rovers [119]; (e) battery cooling [55]. 2 1 Introduction
However, the yield of energy conversion in these materials is still quite low today [cf. 122, 158]. Moreover, the materials with the highest thermoelectric efficiency are in most cases not environmentally friendly and relatively expensive [98]. These facts limit the thermoelectric energy conversion to niche applications. Therefore, it is desirable to develop new, environmentally friendly materials with a good cost-benefit ratio to open up this technology for general public applications. This can be achieved either by cheap and abundant materials or by attaining higher efficiency. Although, thermoelectric materials have been in use for decades for energy conversion, their efficiency has remained at the same level for almost half a century. This was due to a lack of physical insights and new materials. Finally, a new conceptual approach arised in 1993. Hicks and Dresselhaus [70] reported on quantum size effects on the thermoelectric efficiency. Since then, nanostructuring2 has been applied on nanowires [72], phononic nanomesh structures [168], quantum-dot systems [25], nanograined bulk materials [131] and multilayer structures [157] to enhance the thermoelectric efficiency [cf. 122]. Together with material physicist groups around Blochl¨ 3, Jooss4 and Volkert5, we are bound into a priority program of the German research association (DFG SPP 1386). This program was founded to enhance thermoelectric efficiency for power generation from heat through nanostructured materials. As part of this program, we focus on nanoscale structured multilayer and superlattice systems.
2Further information about nanoscale thermoelectrics is given by Pichanusakorn [129]. 3Institute of Theoretical Physics, Clausthal University of Technology 4Institute of Materials Physics, University of Gottingen¨ 5Institute of Materials Physics, University of Gottingen¨
3 1 Introduction
The high potential for multilayer thermoelectric materials has been shown by Venkatasubramanian et al. [156] in 2000. They increased the thermoelectric efficiency significantly by suppressing the thermal conductivity perpendicular to the layered materials.6 Multilayers are amorphous or polycrystalline layered materials. In contrast, in superlattice systems each layer is single crystalline [31]. However, both consist of two different materials, which alternate in a stack. The thickness of each layer material is on the nanometer scale. For such systems the thermal conductivity of the stack cannot simply be predicted from each layer material, because here the transport of thermal energy depends on the phonon propagation across the interfaces and along surfaces at various wavelengths. Hence, the whole package represents a material with new resultant thermal conductivity. Consequently, Fourier’s law of heat conduction [cf. Eq.1.7] would be represented with an effective thermal conductivity κeff. However, to design highly efficient layered thermoelectric materials, it is necessary to understand the influence of specific and intrinsic sample properties onto the anisotropic thermal transport of each layer material and the stack. Therefore, a reliable measurement technique is required to determine the cross-plane and in-plane thermal conductivity of nanoscale samples.
6Although the absolute values are discussed controversially in literature, it indicates a tremendous decrease of the thermal conductivity.
4 1 Introduction
Figure 2: Exemplary collection of thermal conductivity values of different solid materials at room temperature [cf. 71, 106].
Objectives of this work In this work, we mainly focus on the 3ω-method to determine the thermal conductivities of the investigated materials. While several techniques exist to determine the thermal conductivity of solids, thin films or multilayers, the 3ω-method is one of the most well-established due to its high accuracy [78, 126]. This method is especially convenient for poor heat conductors [28], such as thermoelectric materials. In this work, we distinguish the thermal conductivity of the materials to be either poorly-conductive, moderately-conductive or highly-conductive. Since thermoelectric materials are applied with direct currents and relative constant temperature gradients, these terms refer only to the materials thermal conductivity throughout this work. In Fig.2, an overview is given for the thermal conductivity of different solid materials.
5 1 Introduction
However, there is still a demand to further the overall understanding of macroscopic influence factors within a 3ω-measurement in order to determine accurate thermal conductivities. Moreover, no analytic solutions are available for complex measuring configurations, which could allow to measure the anisotropic thermal conductivity in poorly-conductive materials. Therefore, the objectives of this work are: Examine macroscopic influence factors within a 3ω-measurement, develop new geometry configurations to measure the cross-plane and in-plane thermal conductivity, and establish a new methodology, combining experiments and simulations, to identify the isotropic and anisotropic thermal conductivity of materials with nanoscale thickness, such as layered films and pillar geometries.
6 1 Introduction
Outline of the present work In the following introduction chapter, the fundamentals of thermoelectricity are introduced and the materials of major interest are presented. Subsequently, the notion of macroscopic heat conduction in a continuum is derived. Chapter 2 covers the concept and working principle of the 3ω-method. Different analytic solutions of various geometry configurations are compared, the limits are examined and discussed. In chapter 3, the principles of the Finite Element Model are presented including the governing equations, the transient analysis and information about the meshes. In chapter 4, different geometry structures are investigated. First, different macroscopic influence factors are studied for heat sources on top of bulk material configurations and the relevance for a 3ω-measurement is examined. Second, classic monolayer and multilayer configurations are considered. Third, bottom electrode geometries are investigated for bulk-like to thin layer materials. Fourth, two-dimensional and three-dimensional membrane structures are studied. At last, pillar structures are investigated for current induced pulse heating applications. In chapter 5, a new methodology is presented to determine the anisotropic thermal conductivity with Finite Element simulations and Neural Networks. Chapter 6.1 contains two application examples for the thermal conductivity determination with the new presented method. Finally, this work is summarized in chapter 7.
7 1 Introduction
1.1 Thermoelectricity and material selection
Thermoelectricity In general, three distinct thermoelectric effects exist. The first thermoelectric effect was discovered by Thomas Johann Seebeck in 1820-1821, however he dated his memorandum 1822-1823 [142]. The ‘Seebeck-effect’ induces a voltage when two electrically conducting materials in a circuit (thermocouple) have different temperatures
(T1, T2) at their junctions [Fig.3].
Figure 3: ’Seebeck-effect’: Induced voltage in a thermocouple [cf. 10, 137].
The thermo voltage ∆V between the material junctions is defined as the difference of the contact potentials ϕ1/2 [10] or the Seebeck coefficient S times the temperature difference [74, p. 100]
∆V = ϕ ϕ = S (T T ) . (1.1) 1 − 2 · 1 − 2
Shortly afterwards, Jean Charles Athanase Peltier (1834) discovered the reversion of the ‘Seebeck-effect’. Applying a direct current through a thermocouple, consisting of two different materials, leads one intersection to heat up and the other one to cool down [Fig.4] [74, p. 100].7
7The heat produced by the ‘Peltier-effect’ at the up heating intersection surpasses by far
8 1 Introduction
Figure 4: ’Peltier-effect’: Electric current flow in a thermocouple [cf. 74].
The amount of heat per unit time Q, which is absorbed at one intersection and liberated at the other intersection is defined as [138]
Q = Π I . (1.2) ·
Here, Π is the Peltier coefficient and I is the electric current. The ‘Peltier-effect’ is reversible and describes the change in heat content at an intersection between two different materials. The change in heat content results from the flow of electric current across it [137]. The direction of the current flow determines whether the intersection heats up or cools down.8 A connection between the Seebeck coefficient and the Peltier coefficient is given [137] via the temperature by
Π = T S . (1.3) ·
Approximately two decades after the ‘Peltier-effect’ was discovered, William Thomson (1848-1854) observed an additional thermal effect due to the electric current [52]. The ‘Thomson-effect’ describes the fact that heat is either liberated or absorbed within the leg of a thermocouple if an electric current flows through, while a temperature gradient exists.
the heat produced by the ‘Joule-effect’ in the material [63, p. 349]. 8Joule heating takes place in every electrical conductor when an electrical current flows through it and does not require the existence of two different materials. Furthermore, Joule heating is independent of the current’s direction.
9 1 Introduction
(a) (b)
(c) (d)
Figure 5: Working principle of a thermocouple for (a) electric energy generation and (b) cooling application. Fig. (c) shows the arrangement of multiple thermocouples in a module [cf. 145] and Fig. (d) shows a fabricated thermoelectric module [150].
10 1 Introduction
In thermoelectric converters, the thermocouple is rearranged to obtain larger intersections between material A and B [Fig.5(a) and 5(b)]. Thus, the intersections are better exposed for thermal contacts. A thermocouple can be used in two ways. First, applying an external heat input induces a thermo voltage by the ‘Seebeck-effect’. Thus waste heat can be transferred to electric energy. Second, applying an electric current results in cooling of one side of the thermocouple (absorption of heat) and heating up the other side of the thermocouple (heat rejection). Hence, electric energy can be used by the ‘Peltier-effect’ for cooling applications. Since both conversions involve a temperature gradient, the ‘Thomson-effect’ occurs for both applications of the thermoelectric converter, energy harvesting by the ‘Seebeck-effect’ and cooling application by the ‘Peltier-effect’. However, for energy conversion in thermoelectric generators, the ‘Thomson-effect’ is not of primary importance [137]. A key property of the converting modules is the material’s efficiency, described by the ‘figure of merit for thermoelectrics’ (ZT):
σ S2 ZT = ec · T; κ = κ + κ . (1.4) κ · e ph
Here, σec is the electrical conductivity, S the Seebeck coefficient, T the temperature and κ the thermal conductivity. The thermal conductivity is composed of a part due to electrons κe and a part due to phonons
κph [145]. Since every thermoelectric material has a peak performance at a certain temperature [cf. 145], the temperature is included in the ‘figure of merit’ [137]. In general, this peak performance is in the range of 370 1270 K [cf. 145]. Nowadays, industrial applications use − thermoelectric materials of which the maximum ZT 1 [158]. ≈
11 1 Introduction
Figure 6: Replacement of bulk-like material by multilayer structures.
Material selection Consequently, the choice of materials is essential for efficient thermoelectric converters. The conflicting terms of the ‘figure of merit’ need to be optimized in order to increase the ZT value [cf. 145]. However, the distinct terms of the ‘figure of merit’ are connected with each other since all quantities depend on electron properties. Here, the new conceptual approach of Hicks and Dresselhaus [70] could allow the reduction of the phononic part of the thermal conductivity
κph by structuring at the nanoscale and thus decrease the overall thermal conductivity. Normally, each leg of a thermocouple consists of bulk-like materials. These could be replaced by multilayer structures [cf. Fig.6]. For the development of sustainable thermoelectrics, we focus on environmentally friendly oxide materials. The two following materials are of major interest in this work.9 The first material we focus on is the perovskite oxide material strontium titanate SrTiO3 = STO. It consists of strontium, titanium and oxygen. While STO has a cubic crystal structure above 105 K [cf. Fig.7(a)], a structural phase transformation towards a
9Detailed material properties are given in App.B.
12 1 Introduction
(a) (b)